Introduction: An Illusion of "Simplicity" A narrative has recently become common within the team: "Building an Agent is simple now. You can just piece it together with LangChain, BaiLian, or Flowise, ...
引子 从工程师的视角来观察,随着Scaling Law失效问题被更多的人提起,我越来越认同LLM正在逐渐进入「精打细算,收割果实的平庸时代」。Andrew Ng在他的感恩节给读者的来信中提到,AI可能存在泡沫但是一定不是在AI应用开发: AI 应用层: 投资不足。其潜力远超大多数人的认知。 AI 推理基础设施: 仍需大量投资。 AI 模型训练基础设施: 我对这一领域仍持谨慎乐观态度,但可能存在泡沫...
Introduction Viewed through an engineering lens, as the "Scaling Laws" face increasing scrutiny, I find myself agreeing with the growing consensus: Large Language Models (LLMs) are entering a "middle ...